This study presents a machine learning (ML)-based geometry recognition approach to efficiently assess the feasibility of intensified reaction and separation processes. Traditional ternary diagram analysis relies heavily on expert heuristics, making feasibility evaluations time-consuming and subjective. To address this, we integrate k-means clustering-based image segmentation and the Canny edge detection algorithm to automatically extract key graphical features, including reaction equilibrium curves and reachable regions, from ternary diagram. These features are incorporated into tray-by-tray calculations to determine process feasibility under various conditions. The proposed approach is validated through case studies on methyl tert-butyl ether synthesis, 2-pentene metathesis, and tert-amyl methyl ether production. The ML-driven method provides results within seconds, closely matching those from rigorous Aspen Plus simulations. Additionally, it successfully evaluates complex azeotropic systems and double-feed reactive distillation columns, demonstrating its scalability and robustness. By automating graphical interpretation, this method significantly reduces reliance on heuristics and improves design efficiency. The ability to rapidly assess process feasibility enhances decision-making in chemical process development, contributing to energy efficiency and carbon emission reduction. This framework serves as a powerful tool to expedite the feasibility assessment of reactive distillation technologies for their application in azeotropic separation processes.
